Local API server cannot use chat history and local docs

[JamesGai]

This bug report describes the same issue as #3210

When chatting with local docs using API server, it is not aware of what the previous question

I have a frontend UI where users can make conversation with the chatbot. I am using local API server, but discovered a critical bug, the model are not able to provide answers based on what was talked about including the local docs.

Also, the local doc that I have is just a simple text file. You can notice the local doc is retrieved only once when processing the first query regardless query content. However, the following conversation is not able to semantically bind the content from this text file.

I assume each time you send a query to the server, it is a brand new chat session so that's why no history can be preserved. Any solutions?

[matthiasmicheldevelop]

Same problem here using here using GPT4All v3.6.1: Only the first request after startup and enabling the LocalDocs takes the reasoning from the local document. The next request takes it from the model.

[Manon-56]

I have the same problem here. I am testing GPT4all against my own RAG and to do that I provide GPT4all with a series of questions. To automate this, I use the server chat through a python script. I feel like I need to reset the model at each iteration but I can't find the location in the script where it needs to be reset. Plus, it's as if GPT4all app needs to be restarted altogether